Spatiotemporal Multi-Graph Convolution Network for Ride-Hailing Demand Forecasting

论文链接，见这里
作者：Xu Geng, Yaguang Li, Leye Wang, Lingyu Zhang, Qiang Yang, Jieping Ye, Yan Liu
期刊：The Thirty-Third AAAI Conference on Artificial Intelligence (AAAI-19)；2019年
关键字： MGCN,网约车

摘要

区域需求预测是网约车服务的一项基本任务。准确预测网约车需求，可以指导车辆调度，提高车辆利用率，减少等待时间，缓解交通拥堵。由于区域间复杂的时空依赖性，这项任务具有挑战性。现有的方法主要集中在空间相邻区域之间的欧氏相关性建模，而我们观察到可能遥远区域之间的非欧氏成对相关性也对准确预测至关重要。在本文中，我们提出了一种用于网约车需求预测的新型深度学习模型——时空多图卷积网络(ST-MGCN)。我们首先将区域间的非欧几里得成对相关性编码成多个图，然后使用多图卷积对这些相关性进行显式建模。为了利用全局上下文信息来建模时间相关性，我们进一步提出上下文门控递归神经网络，该网络通过上下文感知门控机制来增强递归神经网络，以重新加权不同的历史观测。我们在两个真实世界的大规模网约车需求数据集上评估提出的模型，并观察到与最先进的基线相比有超过10%的持续改进。

论文翻译

《跨时空大尺度网约车流量预测》
《ST-MGCN》

论文视频讲解

小墨STMGCN

代码链接

代码Wang:https://github.com/underdoc-wang/ST-MGCN

MGCN？GCN

from torch import nn
import torch# Static GCN w/ dense adj
class GCN(nn.Module):def __init__(self, K:int, input_dim:int, hidden_dim:int, bias=True, activation=nn.ReLU):super().__init__()self.K = Kself.input_dim = input_dimself.hidden_dim = hidden_dimself.bias = biasself.activation = activation() if activation is not None else Noneself.init_params(n_supports=K)def init_params(self, n_supports:int, b_init=0):self.W = nn.Parameter(torch.empty(n_supports*self.input_dim, self.hidden_dim), requires_grad=True)nn.init.xavier_normal_(self.W)if self.bias:self.b = nn.Parameter(torch.empty(self.hidden_dim), requires_grad=True)nn.init.constant_(self.b, val=b_init)def forward(self, A:torch.Tensor, x:torch.Tensor):'''Batch-wise graph convolution operation on given list of support adj matrices:param A: support adj matrices - torch.Tensor (K, n_nodes, n_nodes):param x: graph feature/signal - torch.Tensor (batch_size, n_nodes, input_dim):return: hidden representation - torch.Tensor (batch_size, n_nodes, hidden_dim)'''assert self.K == A.shape[0]support_list = list()for k in range(self.K):support = torch.einsum('ij,bjp->bip', [A[k,:,:], x])support_list.append(support)support_cat = torch.cat(support_list, dim=-1)output = torch.einsum('bip,pq->biq', [support_cat, self.W])if self.bias:output += self.boutput = self.activation(output) if self.activation is not None else outputreturn outputdef __repr__(self):return self.__class__.__name__ + f'({self.K} * input {self.input_dim} -> hidden {self.hidden_dim})'

[论链]ST-MGCN. Spatiotemporal Multi-Graph Convolution Network for Ride-Hailing Demand Forecasting相关推荐

行人轨迹论文:STUGCN:A Social Spatio-Temporal Unifying Graph Convolutional Network for Trajectory Predictio
STUGCN:A Social Spatio-Temporal Unifying Graph Convolutional Network for Trajectory Prediction用于轨迹预测 ...
论文翻译 SGCN:Sparse Graph Convolution Network for Pedestrian Trajectory Prediction 用于行人轨迹预测的稀疏图卷积网络
SGCN:Sparse Graph Convolution Network for Pedestrian Trajectory Prediction 用于行人轨迹预测的稀疏图卷积网络行人轨迹预测是自 ...
LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation【论文阅读笔记】
文章目录 1.文章为了解决什么问题? 2.文章提出了什么方法? 3.模型架构以及原理? 4.模型有效性分析 5.改进的效果如何以及比较的对象 6.future work 这篇论文提出了LightGCN ...
SIGIR20|LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation
SIGIR20|LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation 文章链接:https:/ ...
行人轨迹论文阅读SSAGCN: Social Soft Attention Graph Convolution Network for Pedestrian Trajectory Prediction
SSAGCN: Social Soft Attention Graph Convolution Network for Pedestrian Trajectory Prediction SSAGCN: ...
论文阅读：LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation
论文阅读:LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation paper:https:// ...
Graph convolution network with similarity awareness and adaptive calibration...MIA,2020.
Song X, Zhou F, Frangi A F, et al. Graph convolution network with similarity awareness and adaptive ...
Graph Convolution Network图卷积网络（二）数据加载与网络结构定义
背景 : 弄懂Graph Convolution Network的pytorch代码如何加载数据并且如何定义网络结构的. 代码地址:https://github.com/tkipf/pygcn 论文地 ...
【论文阅读 SAGLG】Self-Attentive Graph Convolution Network With Latent Group Mining and Collaborative ……
Self-Attentive Graph Convolution Network With Latent Group Mining and Collaborative Filtering for Pe ...
卷积-CNN-GCN-LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation论文笔记
参考博客:https://www.zhihu.com/question/54504471/answer/332657604 1. 卷积连续: ( f ∗ g ) ( n ) = ∫ − ∞ + ∞ ...

[论链]ST-MGCN. Spatiotemporal Multi-Graph Convolution Network for Ride-Hailing Demand Forecasting